1

REINSURANCE BAD DEBT PROVISIONS

FOR GENERAL INSURANCE COMPANIES

JANUARY 2000

Working Party Members: Richard Bulmer (Chairman)Chris GallagherPeter GreenDavid HartPeter MatthewsGaynore MossSimon Sheaf

2

REINSURANCE BAD DEBT PROVISIONSFOR GENERAL INSURANCE COMPANIES

CONTENTS

Section Page

1 Terms of reference 3

2 Theoretical considerations 5

3 A practical approach 8

4 Data requirements 12

5 Use of ratings produced by credit rating agencies 14

6 What if credit ratings are unavailable or out of date? 17

7 Assessment of default probabilities 18

8 Assessment of recovery percentages 22

Appendices

A Select bibliography 24

B Some possible data requirements for the theoretical approach 25

C Comparison of individual agencies' rating categories 31

D Principal factors likely to affect the security of reinsurers 33

E Note by Standard & Poor's on the assessment of credit risk 34

F Default rates and times to default produced by Standard & Poor's 39

G Default probabilities - Moody's 40

H Swiss Re analyses of insurance insolvencies 41

I Examples of paid recovery percentages 45

3

1 TERMS OF REFERENCE

1.1 This paper had its origins in another paper produced by the working party in the specific context of theLloyd's market, which described possible methodologies which could be applied by actuaries who areproviding statements of actuarial opinion on Lloyd's syndicates. Following the completion of theLloyd's paper, the working party was asked to produce a further paper on reinsurance bad debtprovisions which would be of more general application to companies transacting general insurancebusiness.

1.2 There is currently no standard approach to the assessment of reinsurance bad debt provisions.Against this background, the working party has adopted the following terms of reference for thispaper:

§ to prepare a paper on reinsurance bad debts which includes:

§ the principal actuarial issues which need to be considered when setting reinsurance bad debtprovisions

§ possible methodologies which could be applied by actuaries.

1.3 The paper is not intended to prescribe methodologies which actuaries must follow. Working partymembers and the General Insurance Board feel that this would stifle professional judgement.However, the working party would like to encourage a broadly consistent approach among actuariesin the area of reinsurance bad debt provisions. It is considered that such a consistent approach wouldbe beneficial to the standing of the actuarial profession within the general insurance market.

1.4 The scope of reinsurance bad debt provisions normally includes:-

§ currently unpaid reinsurer balances

§ future expected reinsurance recoveries, including recoveries corresponding to gross outstandingclaims, IBNR (incurred but not reported) claims and IBNER (incurred, and reported, but notenough reserved) claims

§ reinsurance disputes as well as reinsurer insolvency

and each of the above elements is discussed in this paper.

4

1.5 The work which can practically be undertaken by actuaries in the area of reinsurance bad debtprovisions for companies transacting general insurance business is often restricted by:-

§ data limitations

§ time constraints

§ budget constraints

A further factor to be taken into account by the actuary when considering how much work toundertake in this area is the level of materiality of reinsurance bad debts relative to the overall level ofclaim provisions.

Consequently, although theoretical considerations are felt by the working party to be very important,the paper concentrates less on what is theoretically ideal but practically impossible, and more on whatis likely to be possible in practice given the constraints listed above.

1.6 Actuaries will need to use their professional judgement to decide how and to what extent theprinciples and methodologies described in this paper should be applied in any individual case.

5

2 THEORETICAL CONSIDERATIONS

2.1 Paragraph 1.5 draws attention to the limitations which frequently restrict the work which canpractically be undertaken by actuaries in the area of reinsurance bad debts. However, it is instructiveto consider the theoretical approach which an actuary might adopt if these limitations did not apply,and the actuary had unlimited data, time and budgets at his/her disposal. We consider it important thatany practical approaches adopted by actuaries in the area of reinsurance bad debts should have asound theoretical basis.

A theoretical approach

2.2 As with other aspects of reserving, there is no single formula or method that will always provideappropriate estimates of future reinsurance bad debts. The approach and assumptions used willdepend on the nature of the inwards business and outwards protections and will incorporate theactuary's experience and judgement. Even if there was a single formula that always providedappropriate estimates, there would probably be other more approximate methods which would beequally reasonable in each individual circumstance.

2.3 The theoretical approach outlined below is a starting point that highlights some of the theoretical issuesinvolved in the analysis and estimation of reinsurance bad debts. The whole theoretical process couldinclude the following stages:-

§ produce an estimate of gross ultimate claims split between:-

§ paid claims

§ reported case estimates

§ IBNR claims

§ IBNER claims

§ allocate the IBNR and IBNER components to individual reported claims, and to an assumedclaims distribution for claims not yet reported

§ produce expected cashflow profiles in respect of each individual claim

§ feed the gross claim projections, including cashflow projections, into a model of the outwardsreinsurance programme, allowing for any currently unpaid reinsurance balances. This results in acashflow profile of reinsurance recoverable for each outwards reinsurance contract split between

§ currently unpaid reinsurance balances

§ reported case estimates

§ IBNR/IBNER claims

This could involve stochastic as well as deterministic modelling

§ for each outward reinsurance contract, identify the reinsurers providing the cover together withtheir percentage shares

§ derive a cash flow profile of expected recoveries from each individual reinsurer

6

§ calculate the probability of recoveries being made from each reinsurer at each future duration.This could be achieved in theory by undertaking a detailed stochastic asset liability modellingexercise for each individual reinsurer

§ combine cash flows and probabilities for each individual reinsurer, and sum for all reinsurers toproduce an estimate of future reinsurance bad debts.

2.4 Net paid claim development triangles produced by some companies only take credit for reinsurancerecoveries to the extent that these have actually been received from reinsurers. The treatment ofoutstanding reinsurance recoveries in net incurred claim development triangles varies from companyto company. In some cases expected reinsurance bad debts are offset against outstandingreinsurance recoveries, while in other cases there is no allowance for expected reinsurance bad debts.If net claim development triangles are adjusted to allow for reinsurance bad debts, then projectedoutstanding claims will also include some implicit allowance for future reinsurance bad debts, butprobably not a sufficient allowance. It will be necessary for the actuary to:

§ obtain an understanding of the treatment of reinsurance bad debts in net claim developmenttriangles provided by the company

§ assess whether or not the treatment of reinsurance bad debts in claim development trianglesintroduces an element of double-counting into the actuary's projections of future reinsurance baddebts. This assessment is likely to involve the application of professional judgement.

2.5 Other theoretical considerations which would need to be reflected in detail include:-

§ principal-to-principal accounting offsets ("set-off"), including allowance for the differentaccounting treatments which apply in different jurisdictions

§ allowance for disputes with reinsurers

§ delays between gross claims being paid and reinsurance recoveries being received

§ dividends from companies in liquidation

§ reinstatement premiums payable when reinsurance recoveries are made

§ bad debt risk for common account protections on inwards proportional treaties, where the natureand security of the common account protections may not be known by the cedant

§ exposures to catastrophe risks. The approach adopted towards reinsurance bad debts in this areaneeds to be consistent with the approach adopted towards gross claim projections.

7

Is the theoretical approach a practical possibility?

2.6 The firm view of working party members is that the answer to this question is usually no.

2.7 The amount of data required to perform the calculations described in paragraph 2.3 would be vast,and would be much greater than the data required to estimate other components of a company's claimprovisions which are likely to be of greater financial significance.

2.8 The time required to perform the calculations would be inconsistent with the tight reporting deadlineswhich are increasingly applicable in the general insurance market.

2.9 The cost involved would be disproportionately large in the context of a typical budget for a year endreserving exercise.

2.10 Perhaps the most compelling argument of all is that the complexity of the theoretical approachdescribed above seems inappropriate in view of the substantial uncertainties involved and the likelydifficulties in interpreting the results of such an exercise. In particular:-

§ the gross claim projections underlying the methodology may be subject to considerable uncertainty

§ the allocation of IBNR/IBNER claims to individual reinsurance contracts will be subject toconsiderable uncertainty due to the range of possible claim distributions which may be fitted tothe historical data

§ the probabilities of reinsurers defaulting will depend substantially on the macroeconomicassumptions which are fed into the stochastic asset liability model. It is difficult to incorporatequalitative factors such as management competence into such a model. However, managementincompetence is a frequent cause of insolvency

§ details of the reinsurers may be unknown or incomplete owing to the role of brokers in placingand processing reinsurances

§ the historic experience of reinsurance insolvencies may not be a reliable guide to futureexperience.

A practical approach

2.11 Having discounted the theoretical approach as being impractical in most situations, it is necessary todetermine a practical approach. However, we consider that it would be unwise to throw away thetheoretical approach entirely. Any practical approach should have a sound theoretical basis.

2.12 We have therefore attempted to construct a practical approach which contains realistic proxies toeach component of the theoretical approach described in paragraphs 2.3 to 2.5, and this practicalapproach is described in Section 3.

8

3 A PRACTICAL APPROACH

3.1 As with the theoretical approach, members of the working party consider that there is no singlepractical approach that will always provide appropriate estimates of future reinsurance bad debts.The precise approach followed will depend on the circumstances of each individual company. Inparticular, the amount of work undertaken by the actuary in the area of reinsurance bad debts willdepend on the materiality of the reinsurance bad debt provision in the context of the overall provisionsof the company.

3.2 The practical approach may include some or all of the following stages:-

§ produce estimates of ultimate claims, gross and net of reinsurance, split between:-

§ paid claims

§ reported outstanding claims

§ IBNR/IBNER claims

§ analyse unpaid reinsurance balances by reinsurer and age of debt. This identifies the currentmajor reinsurance debtors. Investigations into the underlying reasons for slow payment mayprovide insights into the likelihood of bad debts and the presence of disputes on these unpaidbalances and on expected reinsurance recoverables

§ analyse outstanding reinsurance recoveries (corresponding to gross reported outstanding claimsexcluding IBNR/IBNER) by reinsurer

§ obtain a general understanding of the exposure of individual reinsurers to the company'sreinsurance programme and facultative reinsurance arrangements, and the extent ofdiversification between different reinsurers. The following issues, for example, may be ofparticular interest:-

§ are the same reinsurers protecting the property and liability accounts?

§ are the same reinsurers represented on low level working layers and high level catastropheprotections?

§ are the same reinsurers protecting different years of exposure?

§ does the company have a particularly large exposure to one or more individual reinsurers?

§ produce an estimated split of IBNR/IBNER recoveries by reinsurer. The method used to splitsuch recoveries will depend on the general understanding obtained of the company's reinsuranceprotections. For example:-

§ if there is a tendency for the same reinsurers to participate in most protections and in similarproportions, it may be appropriate to apportion IBNR/IBNER claims by reinsurer in the sameproportions as outstanding reinsurance recoveries

§ if different reinsurers protect the property and liability accounts, separate apportionments mayneed to be undertaken in respect of property and liability IBNR/IBNER claims

§ if different sets of reinsurers are on working layers and higher level covers, it may beappropriate to apply different ratios of IBNR/IBNER recoveries to reported outstandingrecoveries for each set of reinsurers

9

§ produce an estimated average duration of future reinsurance recoveries, including IBNR/IBNERrecoveries, for various subdivisions of the company's business. These estimated averagedurations are likely to be derived on an approximate basis from cashflow projectionscorresponding to the estimates of ultimate claims, gross and net of reinsurance, referred to at thestart of paragraph 3.2. Alternatively, the estimated average durations could be derived frombenchmark cashflow projections based on the actuary's wider experience, however the actuarywould need to be satisfied that such benchmarks are broadly appropriate. It may be necessary toproduce separate estimated average durations for various subdivisions of the company'sreinsurance protections. For example:-

§ different classes/groupings of business

§ working layers and higher level protections

§ different years

§ allocate the company's reinsurers to security groupings based on:-

§ credit ratings produced by rating agencies, or

§ other methods where credit ratings are unavailable. The actuary may consider it appropriateto use other methods even when credit ratings are available. The allocation of reinsurers tosecurity groupings is likely to involve the exercise of judgement.

Section 5 of this paper discusses the use of ratings produced by credit rating agencies to groupreinsurers, while Section 6 discusses the approaches which the actuary might follow if creditratings are unavailable or out of date

§ estimate probabilities of groups of reinsurers defaulting on their obligations, based on:-

§ the security of each group

§ the estimated average duration of future recoveries for each group.

It will be necessary to determine separate probabilities for unpaid reinsurance balances andoutstanding reinsurance recoveries. Section 7 of this paper discusses the assessment of thesedefault probabilities

§ estimate the likely percentage recovery from reinsurers who default on their obligations and gointo liquidation. It is usual for significant recoveries to be obtained from such reinsurers, althoughthese recoveries may be delayed very considerably. The assessment of recovery percentages isdiscussed in Section 8 of this paper

10

§ apply default probabilities and assumed percentage non-recoveries (ie 1 - assumed percentagerecoveries) to:

§ unpaid reinsurance balances

§ outstanding reinsurance recoveries including IBNR/IBNER

to produce estimates of future reinsurance bad debts for each grouping of reinsurers.

3.3 Working party members consider that the above practical approach is likely to be appropriate formany companies. However, particular circumstances may require the methodology to be modified toreflect, for example, some or all of the issues identified in paragraphs 2.4 and 2.5. In particular:-

§ the actuary will need to assess whether or not the treatment of reinsurance bad debts in claimdevelopment triangles introduces an element of double-counting into the actuary's projections offuture reinsurance bad debts (see paragraph 2.4)

§ there will be some circumstances in which set-off will be a material factor. This may apply inparticular when a company has a large exposure to one or two individual reinsurers. It may benecessary in such circumstances for the actuary to make an approximate allowance for set-off.The nature of the allowance will depend on the individual circumstances. In some cases, thecompany may be unable to produce a calculation of its full net-of-set-off position. In such cases,it could be argued that the actuary should give no credit, or only limited credit, for set-off.

§ the company may be exposed to one or more disputes with reinsurers. It will be necessary forthe actuary to discuss the nature of the disputes with the company, and to seek an understandingof the likelihood that the company' view will prevail. In some cases, such information may not beavailable to the actuary because the information is sensitive or subject to legal privilege, and thecompany may be concerned about legal issues of discovery. The actuary will then need toconsider whether the outcome of the dispute is material relative to the overall quantum of thecompany's reserves.

3.4 There will also be some circumstances under which the actuary will feel it is necessary to move someof the way from the practical approach towards the theoretical approach. It is for this reason that theworking party has attempted to achieve as much correspondence as possible between the componentsof the theoretical and practical approaches. The circumstances under which the actuary may wish tomove further towards the theoretical approach might include the following:-

§ the company's reinsurance protections are concentrated in a small number of reinsurers

§ the actuary considers that future reinsurance bad debts are likely to be particularly materialrelative to other claim provisions, possibly because the company has a particularly heavy relianceupon reinsurance or because of the current position in the reinsurance pricing cycle

§ the company has a material exposure to catastrophe losses. The issue here is whether or not thecompany's reinsurers are able to survive a major catastrophe

§ the company's reinsurers have a material exposure to Year 2000 losses

Under these circumstances, the actuary may well also wish to undertake some sensitivity testing.

11

3.5 It is envisaged that the actuary will need to use a significant amount of professional judgement whenundertaking many of the stages of the work.

3.6 It will be necessary for estimates of future reinsurance bad debts to be monitored against actualfuture experience.

12

4 DATA REQUIREMENTS

4.1 The actuary's data requirements in respect of reinsurance bad debts will depend on whether he/sheis:-

§ applying some or all of the components of the practical approach described in paragraph 3.2without modification

§ modifying the approach to reflect some or all of the issues identified in paragraphs 2.4 and 2.5

§ moving some of the way from the practical approach towards the theoretical approach asdescribed in paragraph 3.4.

4.2 If the actuary is applying the practical approach without modification or enhancement, the minimumdata requirements would be:-

§ full details of all reinsurance protections, including premium details and names and percentageshares of participating reinsurers

§ unpaid balances split by:-

§ reinsurer

§ age of debt

§ year of origin

§ class of business (where appropriate)

§ currency

§ outstanding reinsurance recoveries, excluding IBNR/IBNER, split by:-

§ reinsurer

§ year of origin

§ class of business (where appropriate)

§ currency

§ details of any disputes with reinsurers.

These requirements are additional to the data required to review other components of the company'sprovisions.

13

4.3 If the actuary modifies the practical approach or moves towards the theoretical approach, datarequirements will become more onerous. For example:-

§ Appendix B contains a detailed data specification which would enable most of the components ofthe theoretical approach, described in Section 2, to be undertaken

§ details of outstanding reinsurance recoveries could be split also by individual reinsurance contract

§ details of the set-off position of monies due to each reinsurer might be requested from thecompany.

4.4 There is a wide variety of possible approaches to the request for data on reinsurance bad debts fromcompanies. These include:-

§ to request only the minimum data required to undertake the practical approach together withwhatever modifications the actuary considers appropriate

§ to request every item of data that the actuary considers might conceivably be relevant.

The working party was not unanimous on the choice of one of these approaches or somewhere inbetween, On balance, members tended towards requesting the minimum data rather than requestingevery conceivable item. It was felt that a company might feel justifiably aggrieved if it produced largevolumes of data that the actuary chose not to use. The amount of data requested may well depend onthe circumstances of the company and the scope of the actuarial exercise to be undertaken, asoutlined in paragraph 3.1.

14

5 USE OF RATINGS PRODUCED BY CREDIT RATING AGENCIES

5.1 The practical approach described in Section 3 includes the allocation of reinsurers to securitygroupings based on credit ratings produced by rating agencies. We consider that, given the limitationsdescribed in paragraph 1.5, this is the most practical methodology available to enable actuaries toassess the security of individual reinsurers. The use of ratings produced by the major credit ratingagencies has the advantages that:-

§ it allows access to a large body of information including historic information concerning theproportions of companies within specific rating groups which have become insolvent withinspecified periods

§ such ratings are objective assessments of a company's security.

5.2 There are a number of different credit rating agencies which produce ratings on insurance andreinsurance companies. The principal of these are A M Best and Standard & Poor's, although ratingsare also produced by Moody's and Duff & Phelps.

5.3 There are some issues arising because of the different rating structures used by the differentagencies. The most appropriate conversion table of which we are aware is provided in Appendix C.Even if a like-for-like rating can be determined from, for example, Bests and Standard & Poor's, therewill be instances where each will rate a company to be subject to different levels of security. In suchcases, a decision would need to be made as to whether the actuary should take a "middle of the road"stance, or take one of the two ratings as being more appropriate. The choice of rating could bedetermined by whether or not there has recently been a favourable or unfavourable trend in thecompany's published credit ratings.

5.4 None of the agencies provides complete coverage of all insurance and reinsurance companiesworldwide, although Standard & Poor's and A M Best provide ratings on virtually all US companies.

5.5 There is a risk that some ratings are only published on payment by the company concerned. Thiscould give a distorted view of the market, as unrated companies will include those which consider theagency to have offered them a rating they consider inappropriately low.

5.6 The ratings are an imperfect measure of security, as they are based on performance tests, the resultsof which have an imperfect correlation with company failure. In spite of this, it is considered that theratings are likely to be the most satisfactory practical means of assessing the risk of failure.

5.7 In general, a full set of ratings is unlikely to be provided more frequently than annually, althoughindividual ratings are updated when specific events occur.

5.8 Where groups of insurance and reinsurance companies exist, the parent company may provide someform of guarantee in respect of recoveries due from subsidiary companies. The value to be placedon such guarantees is a non-trivial issue, but it is considered unlikely that an actuary would be able toprovide a more expert assessment than is already included within the rating agency's assessment.

5.9 In view of the international nature of reinsurance business, it is necessary to compare reinsurers fromdifferent countries with different accounting and regulatory frameworks and using different basecurrencies. It is again considered unlikely that a significant improvement in evaluation could beintroduced by an actuary by comparison with the assessment carried out by the rating agency

15

(including their actuaries) whose full-time function this is. It may, however, be appropriate to make aqualitative adjustment in respect of very recent developments, which may not yet be reflected in thelatest credit agency ratings, although we understand that agencies normally amend their ratings withinone or two months of becoming aware of any new problems.

5.10 There is likely to be some correlation between the failure of one company and others, in that they maybe affected by the same events (e.g. unexpectedly severe catastrophes or a sharp fall in investmentmarkets) and may be reliant on one another for reinsurance recoveries. Consequently, there is thepossibility of a major event causing a "domino effect" and the insolvency of several reinsurancecompanies.

5.11 There may be some finite risk (or other) reinsurance where the reinsurer's basic security is backed bysome form of collateral (e.g. letters of credit or security fund). It would appear appropriate to allowfor this improvement in the security, when ranking reinsurers by security grouping. In the case ofletters of credit, for example, the credit rating of the supporting bank may be used as a guide to theappropriate grouping. However, it should be borne in mind that the collateral would normally coveronly the expected liability and not the exposure.

5.12 When considering the security afforded by Lloyd's syndicates, it may be appropriate to use the creditrating for the Lloyd's market as a whole, in view of the existence of the Central Fund.

5.13 Some companies may be exposed to a very large number of reinsurers. In such cases, it is notunusual to find, for example, that the top 20% of reinsurers represent 80% of the incurred reinsurancerecoveries to date. In such cases, it may be appropriate to allocate the remaining 80% of reinsurersinto a single security grouping or a small number of groupings, depending on the likely level ofmateriality of future reinsurance bad debts.

16

5.14 On occasions, there is a problem of identifying the company where the name of the company mayhave changed or be subject to alternatives in different languages. This may arise from:

§ a simple change of name

§ a change of name to reflect the sale of the company to a new owner

§ the use of the local language and translation into English (Switzerland has particular problems inthis connection in that the name may be in English, French or German)

§ a change of name to distance perceptions of the company from problems or under-performancewhich may have occurred in the past

§ the legal separation of business written in the past from business to be written in the future

§ the reinsurer's participation in reinsurance pools.

17

6 WHAT IF CREDIT RATINGS ARE UNAVAILABLE OR OUT OF DATE?

6.1 In many cases the lack of a current credit rating may indicate that the company concerned is likely toprovide "below average" security, as major reinsurers are likely to be subject to regular rating.However, if the company is new and has not yet reached the stage where the agency believes anappropriate rating can be assessed, the situation is rather different. Nevertheless, it may be that acompany without any track record cannot be considered to be as secure as one with a number ofyears' satisfactory performance.

6.2 In an extreme case, it is conceivable that a reinsurer does not have a rating because it does not exist!

6.3 The materiality of exposure to companies with no current rating needs to be taken into account. Itmay be that detailed investigation or analysis is inappropriate in view of the relative immateriality ofthe exposure to the possible failure of the unrated company.

6.4 If the materiality is significant, some form of grading is still likely to be necessary. This may be basedon information provided by the rating agencies, who analyse the data for many companies even if theydo not provide a credit rating. A list of the principal factors likely to affect the security of reinsurers isprovided in Appendix D.

6.5 A particular issue exists where neither a rating nor the data on which to rate the company is available.The lack of an up-to-date rating may be correlated with the onset of problems with the security of thecompany. The actuary will need to use his/her judgement to form a view, based on whateverinformation is available and taking account of the factors listed in Appendix D.

18

7 ASSESSMENT OF DEFAULT PROBABILITIES

7.1 The probability of a reinsurer going insolvent, and hence defaulting on its obligations, is dependent onmany factors. It will not be possible to incorporate all of these into a practical model, but thepercentage chosen should reflect a true assessment of the underlying risk. The chief factors are:-

§ Financial soundness of the reinsurer.An assessment of this can be made either by using credit ratings published by rating agencies(see Section 5) or by undertaking other investigations when such ratings are unavailable (seeSection 6). We discuss in paragraphs 7.2 to 7.5 how such ratings might be converted intodefault probabilities.

§ Duration of business.Most reinsurers would be on an 'approved' list when reinsurance business is placed, butrecoveries may be made many years after the contract has been written. The longer the periodfrom inception to recovery, the higher the cumulative probability of the reinsurer becominginsolvent. This is illustrated by the table in Appendix G, which shows the cumulative default ratesof bonds with different ratings over several years. These are derived from a study undertaken byMoody's into corporate bond defaults and default rates over the period 1970 to 1998. Althoughthe table is not based on reinsurance defaults, we consider it is still of interest and value. Thetable suggests that the relationship between cumulative default percentages and time is notnecessarily linear. The issue of duration is considered in more detail in paragraphs 7.6 to 7.8.

§ Type of business.The type of business written by the reinsurer would affect the volatility of the results and henceprobabilities of default. Some allowance might be made for this in the factors chosen. However,information may not be available on the reinsurer's mix of business and exposures. Someallowance might be made for this by considering the cedant's book, which is subject to thereinsurance cover, to be a proxy for the reinsurer's book, however care should be taken to ensurethat such an approach is appropriate. For example, catastrophe business requires carefulconsideration as it is a short-tail class but susceptible to substantial variations in claimsexperience. The time when insurers are seeking recoveries from catastrophe reinsurers isexactly the time when the reinsurers' solvency position is most likely to be under pressure.

§ Insurance/reinsurance pricing cycle and economic cycle.These are factors which are common to all or most reinsurers, given that rating trends andeconomic conditions are often strongly correlated across different jurisdictions. History suggeststhat the incidence of insurance/reinsurance insolvencies is correlated closely with the position inthe pricing cycle. This is considered in more detail in paragraphs 7.9 and 7.10.

19

Financial soundness of the reinsurer

7.2 The working party would like to draw attention to four sources of data which may assist actuarieswhen converting credit ratings into default probabilities:-

a note on reinsurance recoverables credit risk which was produced by Standard & Poor's inJanuary 1997 (see Appendix E). This includes details of the derivation from historical data ofdefault rates in respect of reinsurance recoverable credit risk for each rating category. Werecommend that the reader studies the entire note, which was also discussed at a 1997 GeneralInsurance Convention workshop. The working party has been unable to establish the precisedefinition of the "charge factors" mentioned in the note, although we understand they allow forpartial recoveries from liquidated companies.

The default rates are derived from Standard & Poor's database of bond defaults and insurancecompany insolvencies

§ tables produced by Standard & Poor's in January 1998 containing cumulative default rates andtime to default by rating category (see Appendix F). We understand these tables are not specificto the insurance industry. The working party is aware that the January 1998 study has beenupdated by Standard & Poor's, but has not yet been able to obtain a copy of the updated study

§ a study undertaken by Moody's into corporate bond defaults and default rates over the period1970 to 1998 (see Appendix G). Although Moody's only provide a limited coverage of theinsurance sector, the default rates may be applied to equivalent credit ratings used by Bests orStandard & Poor's

§ various graphs published by Swiss Re in their July 1995 Sigma publication showing:-

§ numbers of insurance insolvencies and the insolvency rate during the period 1978 to 1994 inthe US and UK markets

§ the relationship of insurance company insolvencies to profitability, worldwide insuredcatastrophe losses and interest rates.

These graphs are shown in Appendix H. Although the graphs are for the whole market and notfor individual rating categories, they provide a broad reasonableness check on the Standard &Poor's and Moody's analyses, as well as providing an indication of the percentage increase indefault rates which may be expected when general market conditions are adverse.

A possible alternative approach is to derive default probabilities from the differential in yields betweencorporate bonds and risk-free investments for each rating category. Details of yield differentials canbe found in the Bondweek publication (see Bibliography). However, it should be borne in mind thatyield differentials may fluctuate significantly within relatively short time periods, depending on changesin investment conditions. This approach is based on the assumption that the security of corporatebonds can be used as a proxy for the security of reinsurance companies.

20

7.3 It appears that the Standard & Poor's, Moody's and Swiss Re analyses are not obviously inconsistentwith each other, and could therefore form the basis of default percentages. However, it should beborne in mind that such an approach implies that past experience is a reasonable guide to futureexperience. It should also be borne in mind that the Standard & Poor's and Moody's analyses arebased on bond default rates for all market sectors, and not just the insurance market. Furthermore,there may not necessarily be a very strong correspondence between bond defaults andinsurer/reinsurer insolvency. We also suspect that the results of the Standard & Poor's and Moody'sanalyses have been smoothed to some extent. The selection of default percentages is likely to involvethe exercise of judgement.

7.4 Some adjustment to these default percentages may be considered appropriate when there is afavourable or unfavourable trend in credit ratings over the last few years.

7.5 The Standard & Poor's and Moody's studies suggest that the probability of companies rated AAAdefaulting on their obligations during the next 10 years is non-zero. With this in mind, we consider thatthe default probabilities for all reinsurers should be greater than zero, irrespective of the size of thecompany. We appreciate that this will require careful explanation and justification to companymanagement.

Duration of business

7.6 As has been explained above, the longer the period from inception to recovery, the higher thecumulative probability of the reinsurer becoming insolvent. This suggests that different defaultprobabilities should be applied for short-tail and long-tail business.

7.7 The Standard & Poor's, Moody's and Swiss Re data can be used to produce default probabilities foreach rating category.

7.8 These individual year default probabilities can be applied to:-

§ future cashflow projections of reinsurance recoveries for each class of business, or

§ average durations for each class of business. It should be borne in mind that average duration ofoutstanding reinsurance recoveries will vary by year of account, and that the average duration ofreinsurance recoveries may well be longer than the average duration of the corresponding grossclaims, or

§ benchmark average durations based on the actuary's wider experience, however the actuary willneed to be satisfied that such benchmarks are appropriate.

21

Pricing and economic cycles

7.9 Reductions in profitability can lead, albeit with some delay, to an increase in the number ofinsolvencies. The implication would be that when the market is soft this will decrease profitability andhave a knock-on effect on solvency. This might be more important where the recoveries areexpected only over a short period , whereas with longer-tail classes an averaging would take place.Sometimes, when reinsurance rates are soft, reinsurers purchase a larger volume of retrocessionalcover, thus increasing the likelihood of a "domino effect" (see paragraph 5.10) of reinsurancecompany insolvencies. An allowance could be made for soft markets by increasing the defaultprobabilities to allow for this effect.

7.10 Similarly, in the past when interest rates have fallen, insolvencies have tended to increase. Anallowance could be made for this effect. There are two opposing factors at work. If interest ratesrise then assets fall in value, reducing solvency especially if assets have to be sold to pay claims.However high interest rates give high income from investments improving cashflow. If interest ratesfall, cashflow is squeezed and the risk to solvency is increased. It appears from the Swiss Re analysesthat the second effect is more significant.

Other comments

7.11 Care should be taken to avoid double-counting in the assessment of default percentages. It is possibleto apply percentages based on the rating agencies' analyses, and then make additional allowance forknown risks, which may already be reflected in these analyses. Under some circumstances, thiscould result in the over-estimation of future reinsurance bad debts.

7.12 If a company has a particularly large exposure to one or two reinsurers, we consider there is areasonably strong argument that the default percentages attributable to these individual reinsurersshould be increased to reflect the concentration of risk. The assessment of the increase required willlargely be a matter of judgement for the individual actuary.

7.13 It is worth noting that companies can be pro-active in reducing actual default percentages by adoptinga policy of judicious commutation of selected reinsurance contracts.

22

8 ASSESSMENT OF RECOVERY PERCENTAGES

8.1 The fact that a reinsurer becomes insolvent does not usually mean that the creditors will not receiveanything. Ultimately, they are likely to receive a proportion of their claims, often expressed as apercentage or as a number of pence to be received for each pound owed. Failure to allow for thiswhen setting bad debt provisions will cause the provisions to be overstated.

8.2 Obtaining data on ultimate recovery percentages for insolvent reinsurers is difficult. The percentagespaid to date are often publicly available but this is seldom the case with the anticipated ultimatepercentages. A selection of paid recovery percentages is shown in Appendix I. It should be notedthat ultimate recovery percentages will, in many cases, be substantially higher than paid recoverypercentages.

8.3 Clearly, the paid and anticipated ultimate percentages will be virtually identical for any insolventreinsurers whose liabilities are close to being fully run-off. Such companies are likely to have becomeinsolvent some time ago. However, such percentages may still have some relevance to reinsurancebad debt provisions. The underlying reinsurance market regulatory environment and the reasons whycompanies fail are probably broadly similar across the decades, although there may have been someimprovements in regulation in some jurisdictions. We suspect that average ultimate recoverypercentages probably remain of a similar order of magnitude over time.

8.4 Research into a random selection of companies which have become insolvent more recently indicatedthat very few had publicly released any information on their expected ultimate recovery percentages.In addition, where the information is not already in the public domain, the liquidators are unwilling torelease it. Consequently, our limited investigations yielded only two examples of expected ultimaterecovery percentages, which are as follows:

§ Bermuda Fire and Marine: 20% - 30%

§ KWELM: 30% - 50%

8.5 Where an individual syndicate is a creditor of an insolvent reinsurer, it may have access to theexpected ultimate recovery percentage, even if this figure is not in the public domain.

8.6 Historical experience indicates that ultimate recovery percentages tend to be within the range 20% to60%, although there are a few examples which have fallen outside this range.

8.7 A possible alternative approach is to consider average recovery rates as measured by defaulted bondprices. Moody's Investors Service reported in January 1999 that average recovery rates forcorporate bonds, measured on this basis, had fallen to 45% of par from 54% a year previously, butremained above their post-1970 average of 41% of par.

8.8 A recovery percentage needs to be selected for each reinsurer protecting the company under review.For reinsurers which are already insolvent, the percentage selected should, where possible, be basedon information that is in the public domain. However, as discussed above, such information is unlikelyto be available in many cases.

23

8.9 For reinsurers which are not currently insolvent, or those which are insolvent but for which there is nopublicly available information, a percentage will need to be estimated. If there are any reinsurerswhose solvency is particularly significant to the company then it would be sensible to consider themseparately. In all other cases, the best approach would probably be to select average recoverypercentages. Either the same average could be chosen for all reinsurers, or the reinsurers could bedivided into groups and different averages selected for each group. One possible way of grouping thereinsurers would be to use the same groups as for the default probabilities. This may well meangrouping them according to their current credit ratings. On balance, working party members considerthat, unless there are exceptional circumstances, it would be appropriate to use the same assumedrecovery percentage for all solvent reinsurers.

8.10 The recovery percentages could be incorporated into the calculation of the bad debt reserves in one oftwo ways:

a. as a reduction in the probability of default; or

b. as a percentage that will be unrecoverable if default does occur.

Either approach would seem to be perfectly acceptable.

8.11 If it takes a long time for an insolvent reinsurer to pay claims, this will increase the futureadministration costs of the company and could introduce borrowing costs. Obviously, the timing ofrecoveries may be a highly material issue from a commercial perspective.

24

Appendix A: Select bibliography

1 Report of reinsurers security working party by D M Hart et al(1990 GISG).

2 "What ratios really matter?" by J W Dean et al(1993 General Insurance Convention, pages 209 - 236).

3 Workshop session: rating agencies by K Felisky-Watson et al(1995 General Insurance Convention, pages 301 - 310).

4 "Reinsurance Security" by D E A Sanders et al(1996 General Insurance Convention, pages 179 - 204).

5 Workshop session: analysis of reinsurance security by R A Shaw and F J Mackie (1997 GeneralInsurance Convention, page 61). See Appendix E.

6 Swiss Re, Sigma No 7-1995: Development of insolvencies and the importance of security in theinsurance industry.

7 Note by Standard & Poor's on the assessment of credit risk (see Appendix E) - The Review, January1997.

8 Bondweek.

9 Ratings Performance 1997 – Stability and Transition – January 1998 – Standard & Poor's.

10 Historical Default Rates of Corporate Bond Issuers, 1920 - 1998 - January 1999 - Moody's InvestorsService.

25

Appendix B: Some possible data requirements for the theoretical approach

Contents of appendix

B1. Description of data required

B1.1 FGU (from ground up) loss detailsB1.2 Claims bordereaux/statementsB1.2 Event codesB1.4 Unpaid balancesB1.5 Security table

B2. Data table specifications

B2.1 FGU loss details - (not required for proportional contracts)B2.2 Claims bordereaux/statementsB2.3 Event codes - (not required for proportional contracts)B2.4 Unpaid balancesB2.5 Security tableB2.6 Reinsurer security table

B3. Other data

B3.1 Contract detailsB3.2 Premium details

B4. Other contracts

B5. Past disputes

Appendix B cross-refers to paragraph 4.3 of the paper.

26

B1. Description of Data Required

Data specifications for all of the following can be found in sections B2 & B3, and should besubdivided by currency, where appropriate. The following data is in electronic format.

B1.1 FGU (From Ground Up) Loss Details

FGU loss details are required for all loss events, broken down by class, underwriting year andcurrency on both a paid and outstanding basis.

B1.2 Claims Bordereaux / Statements

Detail of paid and outstanding claims by event, contract and section.

B1.3 Event Codes

A complete event code table is required. This table contains for each event code the date of loss anddescription.

B1.4 Unpaid Balances

This information is required in order to produce complete statements for individual reinsurers.

B1.5 Security Table

A security table is required for each contract.

27

B2. Data Table Specifications

B2.1 FGU Loss Details – (Not Required for Proportional Contracts)

DescriptionEvent CodeContract ReferenceSection CodeUnderwriting YearBusiness Type CodeReinsurance Programme Mapping CodePaid FGU ClaimsOutstanding FGU Claims

B2.2 Claims Bordereaux / Statements

DescriptionContract referenceSection CodeCover note referenceBroker codeEvent codeFunds withheldLetters of CreditContribution of claim to agg deductible (Not required for proportional contracts)Paid amount of claimOutstanding claim reserve

B2.3 Event Codes – (Not required for proportional contracts)

DescriptionEvent CodeLoss descriptionDate of lossSpecial claim indicatorType of business code

28

B2.4 Unpaid Balances

DescriptionInternal reinsurer codeBroker codeContract referenceSection codePool indicatorBroker referenceBalance outstanding

B2.5 Security Table

DescriptionContract referenceSection codeInternal reinsurer codeBroker order percentageReinsurer proportion (of order %)Pool codeBroker alpha codeBroker reinsurer referenceLead underwriter indicator

Commuted indicatorPool indicator

B2.6 Reinsurer Security Table

DescriptionInternal reinsurer codeBad debt proportion on unpaid balancesBad debt proportion on O/S reservesBad debt proportion on IBNR reserves*Security rating at Dec `97*Security rating at Dec `96*Security rating at Dec `95*Security rating at Dec `94*Security rating at Dec `93

* Plus any relevant rating agency information i.e. Standard & Poor’s, A M Best.

29

B3. Other Data

The following table definitions are included for completeness

B3.1 Contract Details

Description CommentsContract referenceSection code Used for different contract sections - eg Top &

DropsBroker referenceBroker account codeMain class of businessSub class of businessType of contractContract descriptionUnderwriting referenceMain brokerProportion of whole placementAmount actually placedBroker order percentageCoinsurance percentage if not 100% placedInception date of contractExpiry date of contractExcess point (Not required for proportional contracts)Each and every loss limit (Not required for proportional contracts)Aggregate deductible (Not required for proportional contracts)Aggregate limit (Not required for proportional contracts)Claim rate of exchangePresentation sequence of claimsNo of reinstatements – 1 (Not required for proportional contracts)No of reinstatements – 2 (Not required for proportional contracts)No of reinstatements – 3 (Not required for proportional contracts)No of reinstatements – 4 (Not required for proportional contracts)No of reinstatements - unlimited (Not required for proportional contracts)Cost of reinstatements - 1 (Not required for proportional contracts)Cost of reinstatements - 2 (Not required for proportional contracts)Cost of reinstatements - 3 (Not required for proportional contracts)Cost of reinstatements - 4 (Not required for proportional contracts)Cost of reinstatements - unlimited (Not required for proportional contracts)Reinstatement rate of exchange (Not required for proportional contracts)Security approved

30

B3.2 Premium Details

Description Comments

Contract reference

Section Code

Premium rate on adjustable contracts (Not required for proportional contracts)

Rate of exchange for premiums from slip

Net premium income

Estimated premium income

Flat Premium (Not required for proportional contracts)

Minimum & Deposit premium (Not required for proportional contracts)

Deposit premium (Not required for proportional contracts)

Adjustment premium (Not required for proportional contracts)

Reinstatement rate of exchange (Not required for proportional contracts)

B4. Other Contracts

Any relevant information with regard to facultative covers, surplus covers, financial reinsurances (i.e.time and distance, A.R.T, prospective and retrospective covers etc). In particular, all details inrespect of commutation/sunset clauses.

B5. Past Disputes

Any relevant information with regard to past disputes with reinsurers, which have now been resolved.

31

Appendix C: Comparison of individual agencies' rating categories

Comparison of individual agencies' rating categories

Standard & Poor's1 A M Best2 Moody's1

secure 1 AAA extremelystrong

A++ superior Aaa exceptional

2 AA+, AA, AA- very strong A+ superior Aa1, Aa2, Aa3 excellent

3 A+, A, A- strong A, A- excellent A1, A2, A3 good

4 BBB+, BBB, BBB- good B++, B+ very good Baa1, Baa2, Baa3 adequate

vulnerable 5 BB+, BB, BB- marginal B, B- fair Ba1, Ba2, Ba3 questionable

6 B+, B, B- weak C++, C+ marginal B1, B2, B3 poor

7 CCC very weak C, C- weak Caa very poor

8 R, (U,S)3 extremelyweak

D poor Ca extremelypoor

9 E, F under statesupervision/inliquidation

C

10 S ratingsuspended

11 NR 1- 53 not rated

1) Letters followed by a plus or minus sign (S&P's) and the figures 1, 2 and 3 (Moody's) are notseparate rating categories but indicate whether a company is located in the upper, middle or lowerthird of a rating category.

2) Besides the rating symbols indicated, A.M.Best also uses rating modifiers - letters which giveadditional information on the rating (see below).

3) The figures 1 to 5 indicate why no rating was assigned.

Explanatory notes to Best's Ratings

Not rated categoriesNR-1 Insufficient dataNR-2 Insufficient size and/or operating experienceNR-3 Rating procedure inapplicableNR-4 Company requestNR-5 Not formally followed

32

Rating modifiersg Group Rating, i.e. on basis of consolidated datap Pooled Rating, for companies who pool 100% of their businessr Reinsured Rating, reinsurer's rating when virtually all of the company's business is cededu Under review

Source: A.M.Best

Information taken from "Swiss Re, sigma No. 7/1995", and amended based on comments fromStandard & Poor's and A M Best. It should be noted that any comparison of individual agencies'rating categories involves an element of subjective judgement.

Appendix C cross-refers to paragraph 5.3 of the paper.

33

Appendix D: Principal factors likely to affect the security of reinsurers

Domicile

Currency

Capitalisation and solvency ratios

Profitability, current and projected

Industry trends in premium rates, etc

Growth rate

Changes in mix of business

Management competence and integrity

Strength of technical provisions

Security, adequacy and extent of retrocessional protections

Aggregations of exposure to inwards claims

Quality and spread of assets

Liquidity

Ownership (including any guarantees)

Involvement with intermediaries

Quality of record-keeping and data management

Appendix D cross-refers to paragraph 6.4 of the paper.

34

Appendix E: Note by Standard & Poor's on the assessment of credit risk

The following note was produced by Standard & Poor's in January 1997. It includes details of thederivation from historical data of default rates in respect of reinsurance recoverables credit risk foreach rating category. We recommend that the reader studies the entire note. The working party hasbeen unable to establish the precise definition of the "charge factors" mentioned in the note, althoughwe understand they allow for partial recoveries from liquidated companies.

The ability of reinsurance to help an insurer improve its underwriting results, lower catastrophic risk, bolsterits ability to write new business, and fortify solvency margins and capital position, play an integral role in theevaluation of the financial strength of the insurer. Yet, the ceding of premiums to a vulnerable reinsurer canhave serious implications for an insurer. Indeed, the credit risk associated with reinsurance recoverables isone of the largest risks faced by many property/casualty companies, especially those engaged in significantlonger-tailed liability business.

The analysis of this risk is an area where Standard & Poor's (S&P) can draw upon its worldwide experienceto analyse a primary company's reinsurance portfolio. In fact, of 140 professional global reinsurers, at August1996, S&P had rated 126 companies. Although the reinsurance industry is for the most part secure, S&Pestimates that 4.5% of the US insurance industry's recoverables are at risk (1.02% of aggregated surplus).Moreover, some insurers have a more significant amount of exposure to uncollectible reinsurancerecoverables. This is primarily due to a cursory selection of reinsurers based solely on price or capacity inyears gone by, and is aggravated by a modest capital position.

S&P's capital adequacy model is the primary tool to help evaluate an insurer's capital adequacy. Inrecognition of the reinsurance recoverable risk, the model charges available capital for potentially uncollectiblereinsurance, much as S&P assesses capital charges for investments in bonds with default risk. In the case ofbonds, S&P uses credit ratings of those bonds to gauge default risk, whereas, for reinsurance recoverable,claims-paying ability ratings are used to assess an insurer's vulnerability to reinsurer default risk.

The ability to discern more precisely the impact of a foreign or domestic reinsurer's solvency on a primaryre/insurer's financial strength will improve the ratings process with regard to S&P's evaluation of the capitalstrength of all companies. The US National Association of Insurance Commissioner's risk based capitalformula uses a flat 10% charge to assess an insurer's credit risk from uncollected reinsurance recoverables.The drawback of a flat charge is that a prudent company's true vulnerability to uncollectible reinsurance couldbe erroneously overstated, while a less prudent company's exposure could be understated. By assessingcapital according to the company's reinsurance portfolio, the reinsurance recoverable asset will be modifiedby an allowance for uncollectible reinsurance based on the credit quality of the reinsurer.

35

S&P will be using the following reinsurance recoverable charges, based on the rating of the reinsurer:

Rating of reinsurer Charge factor

AAA, AAAq or AAAISI .005AA, AAq or AAISI .012A, Aq or AISI .019BBB, BBBq or BBBISI .047BB, BBq or BBISI .096B, Bq or BISI .238CCC, CCCq or CCCISI .497NR or UISI .250R or SISI .500

Each charge factor is applied to the total reinsurance recoverable balance collectible from reinsurers rated ineach category. S&P defines the recoverable balance as ceded loss reserves (case and IBNR), lossadjustment expense reserves, unearned premium reserves, and contingent commission reserves, less cededbalances payable, other amounts due to reinsurers, any reinsurer funds held by the insurer, and letters ofcredit. The sum of the products or charge factors times recoverable balances is S&P's estimate of therequired capital to support the company's recoverable risk.

The charge factors are derived from S&P's databases of bond defaults and insurance company insolvencies.Historical analysis of hundreds of corporate bond defaults for the 15-year period ending in 1995 confirms thathigher-rated issues default less frequently than lower-rated issues. For example, the 15-year default rate forbonds rated AAA was 1.4%, while for BBB rated issues the rate was nearly 31%. To apply these corporatedefault rates to insurance insolvency expectations S&P combined the experience of the corporate bond andinsurance markets.

S&P uses a 10-year horizon as the period for which an insurer should hold capital against possiblereinsurance collection problems (the same horizon used in S&P's assessment of default risk on insurers' bondinvestments). Second, S&P correlates the bond default experience with the insurance insolvency experience.S&P then extrapolates expected 10-year insolvency rates for each rating category. Finally, the expectedinsolvency rates are modified to reflect the fact that partial recovery can be made through commutations priorto insolvency or through litigation after the fact. S&P also recognise that insurers may realise slow orreduced payments from weak reinsurers even if insolvency does not occur.

There are limitations that prohibit a perfectly accurate assessment of an insurer's true reinsurancerecoverable risk. Among these are:

§ recoverables for catastrophe reinsurance are reported only when an event occurs, making itimpossible to quantify an insurer's risk to weak catastrophe reinsurers

§ the assessment of the risk is dependent on the insurer's estimate of ceded reserves - if aninsurer's reserves are deficient it is possible that the deficiency also exists in the ceded reserves,and recoverable balances are not available on a line-of-business basis, nor in terms of limits andattachments points for each reinsurer.

36

S&P - Reinsurance recoverables credit risk

S&P Property / Casualty Capital Adequacy Model

The S&P's capital adequacy model is a significant part of the analysis of the capital strength of a Property /Casualty insurer or reinsurer. The model compares total adjusted capital less realistic expectations ofpotential investment losses and credit losses against a base level of surplus appropriate to support ongoingbusiness activities at a secure rating level ("BBB").

This calculation produces a "Capital Adequacy Ratio". An insurer's capital strength is viewed as adequate ifits capital adequacy ratio is at least 100%. The capital adequacy ratio is only a reference point. Qualitativeand quantitative enhancements are applied as warranted to derive a more complete picture.

S&P's Capital Adequacy Ratio

T o t a l a d j C a p i t a l - Asse t - r e l a t e d r is k c h a r g e s - C r e d i t - r e l a t e d r is k c h a r g e s

U n d e r w r i t in g r i s k + R e s e r v e r i s k + O t h e r b u si n e s s r i s k

Evaluation of credit risks

S&P uses Financial Security Ratings of reinsurers to assess an insurer's vulnerability to reinsurer default risk.The reinsurance recoverable charges vary according to the rating of the reinsurer. The NAIC's RBCformula uses a flat 10% charge to assess an insurer's credit risk from uncollected reinsurance recoverables,the disadvantage of this is that a prudent company's true vulnerability would be overstated, with a less prudentcompany's exposure being understated.

Reinsurance recoverables - credit risk charges

Rating Factor Rating Factor

AAA 0.5% S 50%AA 1.2% NR 25%A 1.9% R 50%BBB 4.7% Premiums and agent's balances in course of collection 2%BB 9.6% Premiums and agent's balances not yet due 2%B 23.8% Accrued retrospective premiums 2%CCC 49.7% Interest, dividends and real estate income due & accrued 1%U 25.0% Receivables from parent, subsidiaries and affiliates 5%

Each Charge Factor is applied to the total reinsurance recoverable balance collectible from reinsurers in eachcategory.

37

Recoverable Balance = Ceded Loss Reserve (Case + IBNR)+ Loss Adj Expenses+ Unearned premium Reserves- Ceded balances payable- Other amounts due to reinsurers- Reinsurers funds held by the insurer- Letters of Credit

The sum of the products of charge factors times recoverable balances is S&P's estimate of the requiredcapital to support the company's recoverable risk.

Derivation of reinsurance recoverables charge factors

Bond defaults

These are derived from S&P's database of bond defaults and insurance company insolvencies. Hundreds ofcorporate bond defaults for the 15 year period from 1981 to 1995 were analysed resulting in the calculation ofdefault rates by duration, up to a maximum of 15 years.

Conclusions are:

1 The lower the rating the higher the average cumulative default rate for 1981 - 1995.

15 year default ratesAAA 1.4% BBB 4.7% CCC 45.1%AA 1.5% BB 19.9%A 3.0% B 30.7%

2 The lower the rating the shorter the average time to default.

Time to default (years)AAA 8.0 BBB 6.4 CCC 2.9AA 7.4 BB 5.0A 7.7 B 3.7

Reinsurance recoverables charge factors

§ A 10-year horizon is assumed as the period for which an insurer should hold capital againstpossible reinsurance collection problems.

§ The Bond default experience is then correlated against the insurance solvency experience.

§ Expected 10-year insolvency rates are extrapolated for each category.

§ The expected insolvency rates are modified to reflect that partial recovery can be made throughcommutations or through litigation.

§ Recognition is given to the fact insurers may realise slow or reduced payments from weakreinsurers even if insolvency does not occur.

38

Limitations

The limitations to an accurate assessment of an insurers' true reinsurance recoverable risk are:

§ recoveries for catastrophes are reported only when an event occurs

§ assessment of the risk is dependent on the insurer's estimate of ceded reserves

§ recoverable balances are not available on a line-of-business basis

§ recoverables are not available in terms of limits and attachment points for reinsurers.

Appendix E cross-refers to paragraph 7.2 of the paper.

Source Standard & Poor'sThe Review, January 1997

39

Appendix F: Default rates and times to default produced by Standard & Poor's

Time to Default by Rating Category

Original

rating

Defaults

(units)

Avg. years

from orig. rating

Last rating

prior to D

Defaults

(units)

Avg. years

from last rating

AAA 3 8.0 AAA 0 N.A.AA 9 7.4 AA 0 N.A.A 23 7.6 A 0 N.A.BBB 36 6.6 BBB 7 1.8BB 146 5.1 BB 22 3.1B 233 3.7 B 192 1.9CCC 38 3.2 CCC 267 0.6

Totals 488 4.6 Totals 488 1.3

N.A. – Not applicable

Static Pool Average Cumulative Default Rates (%)

Yr 1 Yr 2 Yr 3 Yr 4 Yr 5 Yr 6 Yr 7 Yr 8 Yr 9 Yr

10

Yr

11

Yr

12

Yr

13

Yr

14

Yr

15

AAA 0.00 0.00 0.06 0.12 0.19 0.35 0.52 0.82 0.93 1.06 1.06 1.06 1.06 1.06 1.06

AA 0.00 0.02 0.10 0.20 0.35 0.53 0.70 0.84 0.91 1.00 1.05 1.11 1.11 1.11 1.11

A 0.05 0.14 0.24 0.39 0.58 0.77 0.97 1.22 1.49 1.76 1.99 2.09 2.17 2.23 2.30

BBB 0.18 0.42 0.67 1.21 1.68 2.18 2.66 3.07 3.38 3.71 3.94 4.06 4.13 4.21 4.21

BB 0.91 2.95 5.15 7.32 9.25 11.22 12.29 13.40 14.33 15.07 16.02 16.44 16.76 16.76 16.76

B 4.74 9.91 14.29 17.42 19.70 21.26 22.56 23.75 24.71 25.55 26.03 26.31 26.43 26.43 26.43

CCC 18.90 26.01 30.99 35.10 39.02 39.88 40.87 41.17 41.86 42.72 42.72 42.72 42.72 42.72 42.72

Inv

grade 0.07 0.18 0.31 0.54 0.78 1.05 1.31 1.58 1.79 2.02 2.19 2.28 2.34 2.38 2.41

Spec

grade 3.75 7.60 11.03 13.79 16.03 17.72 18.90 20.00 20.93 21.73 22.40 22.74 22.96 22.96 22.96

Appendix F cross-refers to paragraph 7.2 of the paper.

Source: Standard & Poor's, Ratings Performance 1997 - Stability & transition – January 1998

40

Appendix G: Default probabilities - Moody's

Average Cumulative Default Rates by Letter Rating from 1 to 20 Years (Percent) - 1970-1988

Years: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Aaa 0.00 0.00 0.00 0.04 0.14 0.24 0.35 0.47 0.61 0.77 0.94 1.13 1.35 1.47 1.59 1.73 1.88 2.05 2.05 2.05

Aa 0.03 0.04 0.09 0.23 0.36 0.50 0.64 0.80 0.91 0.99 1.08 1.18 1.30 1.56 1.63 1.72 1.92 2.04 2.17 2.32

A 0.01 0.06 0.20 0.35 0.50 0.68 0.85 1.05 1.29 1.55 1.81 2.09 2.35 2.56 2.86 3.19 3.52 3.86 4.24 4.45

Baa 0.12 0.38 0.74 1.24 1.67 2.14 2.67 3.20 3.80 4.39 5.04 5.71 6.35 7.02 7.72 8.46 9.19 9.88 10.44 10.89

Ba 1.29 3.60 6.03 8.51 11.10 13.37 15.20 17.14 18.91 20.63 22.50 24.54 26.55 28.21 29.86 31.70 33.28 34.66 35.88 36.99

B 6.47 12.77 18.54 23.32 27.74 31.59 35.04 37.97 40.70 43.91 45.98 47.25 48.53 49.69 51.07 52.20 52.90 52.90 52.90 52.90

Investment-

Grade0.05 0.15 0.33 0.58 0.81 1.06 1.34 1.63 1.94 2.27 2.62 2.99 3.35 3.72 4.10 4.52 4.95 5.36 5.73 5.99

Speculative-

Grade3.82 7.69 11.27 14.44 17.49 20.14 22.33 24.46 26.38 28.32 30.16 31.96 33.74 35.23 36.74 38.36 39.73 40.87 41.87 42.80

All Corp. 1.15 2.30 3.37 4.33 5.21 5.99 6.66 7.31 7.93 8.55 9.17 9.77 10.37 10.91 11.47 12.07 12.65 13.16 13.62 13.98

Appendix G cross-refers to paragraphs 7.1 and 7.2 in the paper. Source: Historical Default Rates of Corporate Bond Issuers – 1920-1998 –January 1999 – Moody's Investors Service.

41

Appendix H: Swiss Re analyses of insurance insolvencies

Annual insurance company insolvency rate

USA

UK

42

Relationship between profitability and insolvencies

USA

UK

43

Relationship of insured catastrophe losses worldwide to number of insolvencies

44

Relationship of number of insolvencies to interest rates

Appendix H cross-refers to paragraph 7.2 in the paper. Information taken from "Swiss Re, SigmaNo 7/1995".

UK

USA

45

Appendix I: Examples of paid recovery percentages

Company Paid recovery percentage declared Date of most recent payment

Bermuda Fire & Marine 1.5% July 1997Bryanston 20.0% 1997Chancellor Insurance 5.0% December 1997Dublin Re 15.0% January 1998English & American 5.0% June 1997Kingscroft 20.0% May 1998Walbrook 13.0% May 1998El Paso 20.0% May 1998ICS Re Private Ltd 50.0% October 1996Lime Street 21.0% May 1998Mentor 70.0% January 1998Mutual Reinsurance 13.0% May 1998Orion 15.0% October 1997RMCA 50.0% August 1996Scan Re 12.5%Trinity 35.0% 1997United Re 40.0% January 1997

Appendix I cross-refers to paragraph 8.2 in the paper.

JRB\H:\Private\jrb\INSTITUT\39R4a.doc

REINSURANCE BAD DEBT PROVISIONS

SUPPLEMENTARY ADVISORY NOTE

OCTOBER 2005

46

Introduction

1 An Advisory Note was issued by the General Insurance Board in the autumn of 1999 entitled Reinsurance Bad Debt Provisions for Lloyd s Syndicates . The Advisory Note described

possible methodologies which could be applied by actuaries who are providing statements of actuarial opinion in respect of Lloyd s syndicates.

2 At that time, there was no standard approach to the assessment of reinsurance bad debt provisions. Against that background, a working party was established with the following terms of reference:

To prepare a paper on reinsurance bad debts which included:

The principal actuarial issues which need to be considered when setting reinsurance bad debt provisions

Possible methodologies which could be applied by actuaries.

3 The original Advisory Note was not intended to prescribe methodologies and factors which actuaries must follow. It was felt that this would stifle professional judgement. However, the profession wanted to encourage a broadly consistent approach among actuaries in the area of reinsurance bad debt provisions. It was considered that such a consistent approach would be beneficial to the standing of the actuarial profession within the general insurance market.

Does the Advisory Note need to be updated?

4 The original Advisory Note is now six years old and it is appropriate to consider if there are any aspects of the paper which need to be updated.

5 Following consideration of the Advisory Note, the General Insurance Board has concluded that:

The principles and overall approach outlined in the original Advisory Note remain generally applicable, and there is no need for it to be re-written

However, there are a number of specific tables in the original Advisory Note which should be updated.

This supplementary advisory note

6 It has been decided therefore to issue this Supplementary Advisory Note which contains updates to three tables from the original Advisory Note. These updated tables are provided with the intention that they should not be used mechanistically, but in a manner consistent with paragraph 3 above.

7 The updated tables in the Appendices to this supplementary guidance note are:

Default rates and times to default produced by Standard and Poor s (See Appendix F of the Advisory Note)

Default probabilities - Moody s (See Appendix G of the Advisory Note)

Examples of paid recovery percentages (See Appendix I of the Advisory Note)

These tables are included in this Supplementary Advisory Note with the permission of Standard & Poor s, Moody s and PricewaterhouseCoopers respectively.

47

Appendix 1

Default rates and times to default produced by Standard & Poor s

Time to Default by Rating Catergory

Original rating Defaults Average Years

from Original Rating

Last Rating prior to D

Defaults Average Years from Prior

Rating AAA 3 8.0 AAA 0 N.A. AA 18 12.0 AA 0 N.A. A 54 11.8 A 0 N.A. BBB 127 7.9 BBB 8 0.6 BB 383 5.8 BB 28 1.6 B 766 4.3 B 289 1.2 CCC/C 64 2.8 CCC/C 817 0.4 N.R. N.A. N.A. N.R. 273 3.7 Total 1,415 5.4 Total 1,415 1.2

N.A. - Not applicable

N.R. - Removed Cumulative Average Default Rates 1981 to 2004 (%)

Rating Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y10 Y11 Y12 Y13 Y14 Y15

AAA 0.00 0.00 0.04 0.07 0.12 0.21 0.31 0.48 0.54 0.62 0.62 0.62 0.62 0.62 0.62

AA 0.01 0.03 0.08 0.16 0.26 0.40 0.56 0.71 0.83 0.97 1.09 1.23 1.36 1.50 1.61

A 0.04 0.13 0.26 0.43 0.66 0.90 1.16 1.41 1.71 2.01 2.24 2.44 2.64 2.81 3.08

BBB 0.29 0.86 1.48 2.37 3.25 4.15 4.88 5.60 6.21 6.95 7.69 8.32 9.01 9.81 10.67

BB 1.28 3.96 7.32 10.51 13.36 16.32 18.84 21.11 23.22 24.84 26.50 27.84 29.08 29.93 30.94

B 6.24 14.33 21.57 27.47 31.87 35.47 38.71 41.69 43.92 46.27 48.19 49.87 51.41 53.24 54.73

CCC/C 32.35 42.35 48.66 53.65 59.49 62.19 63.37 64.10 67.78 70.80 70.80 70.80 70.80 72.26 72.26

Investment Grade

0.11 0.32 0.57 0.90 1.25 1.62 1.95 2.27 2.57 2.89 3.18 3.43 3.68 3.95 4.27

Speculative Grade

5.32 10.88 16.15 20.61 24.24 27.43 30.18 32.65 34.82 36.77 38.44 39.84 41.13 42.43 43.58

All Rated 1.70 3.44 5.03 6.40 7.50 8.45 9.24 9.94 10.55 11.12 11.61 12.01 12.40 12.79 13.20

Source: Standard & Poor s Global Fixed Income Research

48

Appendix 2- Default probabilities - Moody s

Average Issuer-Weighted Cumulative Default Rates by Whole Letter Rating, 1920-2004 Cohort Time Horizon (Years)

Rating 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Aaa 0.00 0.00 0.02 0.09 0.19 0.30 0.41 0.59 0.77 1.01 1.22 1.37 1.57 1.66 1.70 1.80 1.90 1.95 2.07 2.13

Aa 0.06 0.19 0.32 0.49 0.78 1.11 1.48 1.85 2.20 2.57 3.01 3.50 3.98 4.48 4.87 5.13 5.35 5.57 5.87 6.09

A 0.08 0.25 0.54 0.87 1.22 1.58 1.98 2.34 2.76 3.22 3.71 4.21 4.65 5.09 5.56 6.02 6.30 6.60 6.89 7.19

Baa 0.31 0.93 1.69 2.55 3.40 4.28 5.12 5.95 6.83 7.63 8.42 9.22 10.00 10.70 11.32 11.91 12.51 13.04 13.49 13.95

Ba 1.39 3.36 5.48 7.71 9.93 12.01 13.84 15.65 17.25 19.00 20.60 22.16 23.72 25.10 26.31 27.44 28.59 29.70 30.58 31.48

B 4.56 9.97 15.24 19.85 23.80 27.13 30.16 32.62 34.74 36.51 38.24 39.80 41.23 42.67 43.92 45.21 46.15 46.89 47.52 47.79

Caa-C 15.07 24.77 31.82 36.76 40.50 43.63 45.85 47.94 49.89 51.64 53.63 55.61 57.33 59.19 60.94 62.52 63.90 65.34 66.69 67.95 Investment Grade

0.15 0.47 0.88 1.34 1.84 2.35 2.88 3.39 3.93 4.47 5.04 5.61 6.15 6.67 7.14 7.57 7.93 8.27 8.60 8.93

Speculative Grade

3.83 7.78 11.49 14.80 17.73 20.29 22.52 24.53 26.29 28.02 29.67 31.24 32.77 34.19 35.44 36.65 37.76 38.81 39.67 40.45

All Rated 1.48 3.07 4.59 5.99 7.24 8.37 9.38 10.30 11.16 12.01 12.84 13.65 14.43 15.15 15.80 16.41 16.93 17.43 17.88 18.30

Source: Moody's

49

Appendix 3

Examples of paid recovery percentages (London Market)

Run-off Scheme of Arrangement Current Dividend BAI (Run-off) 5% Chester Street 5% Drake 52% English & American 30% ICS (UK) 100% Monument 36.9% (final) OIC Run-off (formerly Orion), London & Overseas 45% Paramount 40% Sovereign 40% Valuation Scheme of Arrangement

Andrew Weir (converted from run-off) 49.65% (final) Anglo American (converted from run-off) 90% Aneco 72.6% Bermuda Fire & Marine (converted from run-off) 40% Black Sea & Baltic 30% BNIB 100% (final) Bristol Re 49% Bryanston (converted from run-off) 39% Chancellor (converted from run-off) 41% Charter Re 90.58% Compagnie Europeenne de Reassurances Undeclared Fremont (UK) 38.3% (final) Hawk 23% (final) ICS Re 88.8% (final) KWELM (converted from run-off) K-65%, W-65%, E-72%, L-68%, M-53% Marina Mutual Insurance Association Limited Undeclared Municipal General 60% North Atlantic Undeclared Pan Atlantic Undeclared Pine Top 24.9% (final) RMCA Re 93% (final) Scan Re (converted from run-off) 80.5% (final) Stockholm Re (Bermuda) 36.376% (final) Taisei Fire & Marine Insurance Company 62% Trinity (converted from run-off) 67.5% United Standard 27% Liquidation

National Employers Mutual General ( NEMGIA ) 37.58% (final)

Source: PricewaterhouseCoopers.

Note: Some of the above percentages are expected to increase further over time.

50